<aside> <img src="/icons/circle-dashed_purple.svg" alt="/icons/circle-dashed_purple.svg" width="40px" />

"DeepSeek has beaten OpenAI"

</aside>

Reality

• Explanation:
  • First there’s a huge difference between optimizing for performance and optimizing for efficiency.
    • It’s like you can spend 2 hours at the gym to maximize muscle gain (optimize for performance) OR you spend 45 minutes to achieve 80% of the gains (optimize for efficiency).
  • While DeepSeek’s reasoning model R1 matches OpenAI's reasoning model o1 performance, OpenAI has already demonstrated o3, which is more powerful than o1 and R1. An oversimplified answer here is DeepSeek leads in efficiency but not overall capability.
    • Note: OpenAI just released o3 mini for free users, and we can use o3 mini with search, which might not have happened without DeepSeek.
• Analogy: It’s like if you have the money, you can pay 1000 dollars for the latest flagship iPhone, but let’s say a Chinese company produces a smartphone that gives you 90% of the flagship performance for 200 dollars. A total hypothetical of course.
  • And although that 200 dollar smartphone might be extremely popular, we can’t say the more cost-effective smartphone beats the iPhone in overall capability.